Also known as: Similar Audiences, Lookalike Modeling, Act-Alike Audience
A lookalike audience is a targeting segment an ad platform builds by finding new users whose traits and behavior statistically resemble a source list of your existing customers. You supply the seed (converted traders); the platform's model finds strangers who look like them.
The mechanism is pattern-matching at scale. You upload a first-party list — hashed emails or phone numbers of traders who deposited — and the platform compares that seed against its entire user graph across thousands of signals: interests, device, purchase history, page behavior, demographics. It then ranks the wider population by similarity to your seed and lets you target the closest slice.
Size controls the trade-off between reach and precision. On Meta, a 1% lookalike targets the ~2.1 million people (in a country the size of the US) who most resemble your seed; a 10% lookalike widens to roughly 21 million but dilutes similarity. For Forex acquisition, tighter percentages usually convert better because trading intent is rare in the general population.
Quality is bounded entirely by the seed. A lookalike built from 500 funded traders who each deposited over $1,000 produces a fundamentally better model than one built from 5,000 people who merely clicked a banner. Garbage seed, garbage audience — the algorithm faithfully replicates whatever you feed it.
You export a customer list from your CRM — ideally your highest-value segment — and upload it to the ad platform, which hashes the identifiers (SHA-256) before matching so raw data is never exposed. The platform matches your hashes against its own hashed user records to reconstruct the seed inside its graph.
A machine-learning model then profiles what your seed users have in common and scores the remaining population by resemblance. You choose a similarity band (1%, 3%, 5%, up to 10%) and the platform serves your ads to that band. Because trading conversions are a strong, specific signal, feeding conversion events back via the pixel or Conversions API continuously sharpens later lookalikes.
Segment your CRM for funded, active traders — not clickers. Aim for at least 1,000 records so the model has enough signal.
Upload emails/phones as a Custom Audience; the platform hashes them client-side and matches against its graph.
Pick a percentage (start at 1–2%) and a country. Narrower bands mean higher resemblance, lower reach.
Optionally add age or exclusion filters, but avoid over-constraining — the lookalike already encodes intent.
Send deposit events via pixel/CAPI so the next refresh of the audience is trained on real outcomes.
Why it matters for partnership: Lookalikes let an IB scale acquisition volume without collapsing lead quality — you clone your best depositors instead of guessing at cold interests. That keeps CPA down and deposit rates in the range brokers pay premium tiers for.
An IB for IC Markets exports 1,400 traders who each deposited over $1,000, hashes and uploads them to Meta Ads Manager, and builds a 1% lookalike across Tier-1 English markets. Running the same creative to that lookalike versus a broad interest audience, the IB sees cost per registration drop from roughly $18 to $11 and the deposit rate on those registrations roughly double over a 30-day test.
| Audience type | Source | Intent | Best use |
|---|---|---|---|
| Lookalike | Model of your converters | Inferred, high | Scaling cold reach that resembles buyers |
| Interest | Platform interest tags | Broad, unproven | Top-of-funnel discovery, cheap tests |
| Retargeting | Your site visitors | Warm, known | Closing people who already engaged |
Build your lookalike from your top decile by deposit value, not your whole list — the model clones whatever you feed it, so feed it your best.
Seeding a lookalike from a small or unqualified list (mere clickers or a few dozen records) produces a fuzzy, low-intent audience and quietly burns your media budget.
Platforms match a minimum of around 100 users, but 1,000 or more highly qualified records gives the model enough signal to find reliable patterns.
Identifiers are hashed before matching so the platform never sees raw data, but you still need a lawful basis and consent under GDPR/CCPA to use customer data for ad targeting.
Begin at 1–2% for the tightest resemblance and best conversion quality, then widen only if you need more reach and can accept a lower deposit rate.
Rebuild every few weeks or when your best-customer profile shifts, so the model tracks who is actually converting now rather than months ago.
Yes, but layer lightly — heavy extra filters shrink and distort the audience, since the lookalike already encodes the traits that matter.
They are allowed on major platforms, but trading and CFD ads face extra ad-policy restrictions and eligibility checks, so review each platform's financial-services rules before scaling.